Your Search for Unrecorded Liabilities Is the Best AI Task on the Close Checklist, and Almost Nobody Posts a Job for It
Lorem ipsum dolor sit amet, consectetur adipiscing elit. Donec rhoncus neque sed nibh sagittis, fringilla porttitor ipsum tristique. Nulla interdum ex et nisi vehicula, id laoreet nisl ultricies. Phasellus vitae magna ac lacus dictum tincidunt. Sed iaculis metus nec viverra pulvinar. Etiam id nisi eu turpis mattis imperdiet ac ac tortor. Aliquam at ipsum dui. Etiam pharetra consequat massa. Aenean nec lectus sit amet metus pharetra dapibus. Pellentesque interdum ex eget nisi fringilla, id semper erat rhoncus. Suspendisse lectus leo, malesuada pharetra commodo a, sollicitudin eu erat. Nullam justo nisl, tincidunt vel auctor id, luctus a tellus.
Aliquam convallis condimentum volutpat
Pellentesque sollicitudin mauris sit amet enim volutpat, at faucibus sem laoreet. Morbi egestas ex non orci interdum, ut elementum orci faucibus. Maecenas et sem convallis erat dignissim facilisis. Quisque purus sapien, pellentesque euismod varius id, fermentum nec nibh. Integer commodo dignissim ipsum, ac accumsan metus fringilla sit amet. Aenean aliquam sem finibus tempor venenatis. Aliquam ac facilisis turpis, eu posuere ipsum.
In nisi dui, ultricies sit amet gravida vel, ullamcorper vel mauris. Aliquam nec sapien odio. Sed vitae suscipit felis. Nullam semper blandit lectus, eu finibus urna fermentum et. Aliquam vehicula ligula nibh, non efficitur massa iaculis et. Vestibulum vitae euismod odio, non maximus nulla. Sed viverra porta enim ac interdum. Maecenas auctor tristique auctor. Nullam et neque nec tortor malesuada ullamcorper. Pellentesque ac fringilla ante, non convallis est. Proin velit augue, rutrum vitae ipsum vel, malesuada dictum urna. Nunc vulputate sit amet odio vitae ullamcorper. Nullam suscipit ornare eros, et viverra sapien hendrerit quis. Donec odio eros, ultricies a risus quis, efficitur elementum turpis. Etiam interdum diam quis turpis ultricies.
Sed euismod quam vestibulum
Sed non sapien eros. Duis fringilla fringilla lectus sit amet aliquam. Aliquam erat volutpat. Vivamus molestie, felis rutrum luctus pulvinar, libero metus eleifend mauris, semper malesuada ante eros vitae eros. Phasellus vitae dolor faucibus, laoreet lectus quis, placerat nisi. Nam ornare nulla id est aliquet, quis fringilla neque congue. Duis facilisis sed massa vel bibendum. Curabitur sollicitudin tristique commodo. Vivamus facilisis venenatis nibh. Integer placerat elementum felis, id consequat lorem consectetur a. Duis laoreet sit amet nisl in eleifend. Interdum et malesuada fames ac ante ipsum primis in faucibus.
Proin eros lacus, pellentesque sed vehicula a, luctus non nibh. Nulla diam sem, posuere ac odio varius, ultrices tristique nibh. Morbi dictum scelerisque convallis. Praesent faucibus lorem lacus, id luctus justo feugiat et. Curabitur eget tellus non nisi interdum blandit. Maecenas pulvinar est sed ex elementum, ac commodo diam bibendum. Nulla auctor dolor felis, sit amet euismod ante eleifend non. Donec id neque magna.
Of the 2,464 accounts payable postings in our database right now, 6 mention unrecorded liabilities. Six. And in the same set, 499 mention accruals, which is the same problem from the other side, so the profession clearly does the work and mostly doesn't name it. I think that makes the search for unrecorded liabilities one of the best places a controller can put AI this quarter, because it's tedious, it's mostly reading, and your auditors are going to do it to you anyway.
What is the search for unrecorded liabilities?
It's the check that asks whether every bill for goods or services you received before period end made it into the books by close. The classic version has you pull disbursements and invoices dated after period end, look at what was paid or received in the first few weeks of the next period, and trace each one back to see whether the service actually belonged to the prior period. Auditors lean on it for completeness of payables, and there's a decent walkthrough of the procedure in this CPA study guide on payables and unrecorded liabilities.
Doing it by hand means a few hundred invoices, a spreadsheet, and a lot of squinting at service dates, and that's where I'd put AI first.
Why does AI fit this check so well?
Because every step is a comparison. You have an invoice with a service period, you have a posting period, and you have a question about whether they line up. A model can read a few hundred post-close invoices in the time it takes you to open the folder, pull the service dates out of the PDFs, and sort them into three piles: clearly next period, clearly last period and needs a human. The middle pile is small once it's working, and that's the part you review.
The second place it helps is the gap check. Give it twelve months of history by vendor and ask which recurring vendors have no invoice in the current period, and you've got a list of likely missing accruals in minutes, which is the thing people forget about because there's no document to look at. A utility, a SaaS tool, a contractor on a monthly retainer. No invoice yet doesn't mean no liability, and a model comparing patterns finds those faster than a person scrolling a vendor ledger.
How do you set it up without losing the audit trail?
Treat the AI as a preparer, and keep every output tied to a source document. Give it a defined input set, which is the AP subledger entries dated after close, the post-close invoice folder, open purchase orders and receiving records if you have them, and a schedule of recurring vendors. Ask it for a table with one row per item that carries the vendor, invoice number, amount, service period, a proposed period assignment and a link or file reference to the source. Then have someone with authority approve each proposed accrual before it posts, and keep the table as your support.
That table is also what you hand your auditors, which is where this saves you real time. If you've already done the search and documented it, the audit request becomes a file transfer instead of a fire drill, and we covered the broader version of that in our piece on audit prep and PBC requests.
What can AI not do here?
It can't know about a verbal commitment, a delivery that sat on a dock, or a vendor who hasn't billed because the project isn't finished. I'd say that's the honest limit and it's a big one, since the riskiest unrecorded liabilities are usually the ones with no invoice at all. The model finds the pattern gaps and the misdated invoices, and the people closest to operations find the rest, so one good habit is sending operations a short list each close and asking them what was received but not billed. If you've been burned by this before, you know the bullshit excuse is always that nobody told accounting.
There's also the risk that the model reads a service period wrong, which is why the unclear pile exists and why you sample the clear piles instead of trusting them. A control you sampled and documented is better than one you assumed, and it stays that way whether the model is right or wrong.
What do the job postings say about where AP is headed?
In the AP roles we track, 437 of 2,464 postings, about 18%, carry an automation signal, 113 or roughly 5% reference AI, and 151, about 6%, name an AP automation tool. As of October 2026 that means most AP teams are hiring for the old workflow with a little automation on the side, and the people who can run both are rare. That's an opening, and our guide on where to start with AI in accounts payable and the vendor statement reconciliation post are the companion pieces to read next. If you're shopping for tools, the Audit Friendly software directory lists the AP and close platforms side by side, and if you want to see who's hiring for these skills, our remote accounting jobs page is updated daily.
What I'd do this close
Pick last month, export the post-close invoices and the vendor history, and run the whole exercise as a dry run next to your normal process. Compare what the model flagged against what you booked, count the misses on both sides and look at why. Honestly the first pass will be messy, and that's fine, because the value is in learning where it's wrong before you rely on it. Then run it live next close with your approval step in place. Build something small and imperfect now and you'll have a real workflow by the time the auditors show up.
Frequently asked questions
What is a search for unrecorded liabilities?
It's a check of invoices and disbursements dated after period end to find costs that belong to the prior period but were never booked, which supports the completeness of accounts payable.
Can AI replace the search for unrecorded liabilities?
It can do most of the reading, date matching and gap detection, but a person still has to approve proposed accruals and chase items that have no invoice yet.
What data should I give an AI tool for this?
Post-close AP entries, the post-close invoice folder, open purchase orders and receiving records, and a history of recurring vendors, with every output tied back to a source document.
Will auditors accept AI-prepared support?
Auditors care about whether the evidence is complete and traceable, so a documented table with source references and human approval gives them what they need, though you should confirm expectations with your audit team first.
How many accounts payable jobs mention AI or automation?
In the 2,464 AP postings we track, about 5% reference AI and about 18% carry an automation signal, as of October 2026.
